[MRG + 2] EHN additional test for trees regarding fitting behaviour with constant features #8580
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
This seems like a good unit test, but can you explain what the issue is a bit more? What do you mean by "the other splitter"? |
|
I am referring to a tree growth using a depth-first approach. Each node
at a given depth will be given to a splitter object. If for the first
node, the splitter find a constant feature, it should mark this feature
as constant for subsequent deeper (i.e. next depth-level)
nodes. However, the other nodes at the same depth should not be affected
by this assignation.
I did this mistake in the other implementation by invalidating the
feature for all nodes (deeper and the current depth level) and the test
did not fail. With the following data, you should have the following
behaviour:
* a first split without any issue;
* at depth=1, one of the node will face
a constant feature while the second one will not face this case;
* at depth=2, only one of the node will be split.
With the bad implementation, you will stop at depth=1. Just to make it
clear, there is nothing wrong with the current implementation. This is
just an additional test to avoid a future mis-implementation.
Hope that this is more clear.
|
|
Got it, this looks good to me. Thanks for the contribution. |
raghavrv
reviewed
Mar 13, 2017
sklearn/tree/tests/test_tree.py
Outdated
| assert_equal(est.tree_.max_depth, 2) | ||
| assert_equal(est.tree_.node_count, 5) | ||
|
|
||
| for name, TreeEstimator in REG_TREES.items(): |
|
Oopsy, I was not looking at the right graph. It make more sense now ;) Not sure it makes sense to test this feature, for the ExtraTree. |
225995f to
e42b504
Compare
|
Thx. +1 for merge from me as well... Will merge when appveyor gives a go.. |
|
Too slow. |
Closed
herilalaina
pushed a commit
to herilalaina/scikit-learn
that referenced
this pull request
Mar 26, 2017
…ith constant features (scikit-learn#8580) * TST add test checking the behaviour of constant/no-constant features * FIX/TST factorize test * TST Add additional constant features * FIX/TST remove ExtraTree from test
massich
pushed a commit
to massich/scikit-learn
that referenced
this pull request
Apr 26, 2017
…ith constant features (scikit-learn#8580) * TST add test checking the behaviour of constant/no-constant features * FIX/TST factorize test * TST Add additional constant features * FIX/TST remove ExtraTree from test
Sundrique
pushed a commit
to Sundrique/scikit-learn
that referenced
this pull request
Jun 14, 2017
…ith constant features (scikit-learn#8580) * TST add test checking the behaviour of constant/no-constant features * FIX/TST factorize test * TST Add additional constant features * FIX/TST remove ExtraTree from test
NelleV
pushed a commit
to NelleV/scikit-learn
that referenced
this pull request
Aug 11, 2017
…ith constant features (scikit-learn#8580) * TST add test checking the behaviour of constant/no-constant features * FIX/TST factorize test * TST Add additional constant features * FIX/TST remove ExtraTree from test
paulha
pushed a commit
to paulha/scikit-learn
that referenced
this pull request
Aug 19, 2017
…ith constant features (scikit-learn#8580) * TST add test checking the behaviour of constant/no-constant features * FIX/TST factorize test * TST Add additional constant features * FIX/TST remove ExtraTree from test
maskani-moh
pushed a commit
to maskani-moh/scikit-learn
that referenced
this pull request
Nov 15, 2017
…ith constant features (scikit-learn#8580) * TST add test checking the behaviour of constant/no-constant features * FIX/TST factorize test * TST Add additional constant features * FIX/TST remove ExtraTree from test
jwjohnson314
pushed a commit
to jwjohnson314/scikit-learn
that referenced
this pull request
Dec 18, 2017
…ith constant features (scikit-learn#8580) * TST add test checking the behaviour of constant/no-constant features * FIX/TST factorize test * TST Add additional constant features * FIX/TST remove ExtraTree from test
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Reference Issue
Related to #8231
What does this implement/fix? Explain your changes.
In #8231, the following error in the implementation did not trigger any failing in the tests of the trees:
When a feature was found to be constant at a given splitter, this feature was declared constant for the remaining of the fitting. However, this is a wrong behaviour since this feature is potentially non-constant for the other splitter.
This additional test, check that the number of nodes and the tree depth are correct with a feature which would be declared constant at the first split as described previously.
Any other comments?